# [R-sig-ME] FW: Prediction of random effects - logist mixed model

Correa S.T. scorrea at soton.ac.uk
Wed May 14 13:03:51 CEST 2008

```
Dear list,

I am working with a two-level logistic mixed model and I am interested
in predicting the random effects for a given value of the parameters
(not for the estimates obtained from the data at hand, which can be
obtaiend using fucntion 'ranef'). For illustration, please see the code
below.

# mixed logistic model

fit1.b<-lmer(yij.b.true  ~ x1 + x2 + (1|area), family =
,control=list(usePQL=FALSE),verbose=FALSE,method="Laplace")

# estimates of the parameter based on the data at hand

beta.hat<-fixef(fit1.b)
varu.hat<-as.numeric(VarCorr(fit1.b)[][1,1])

# predicting group effects based on the data at hand

u.pred<-ranef(fit1.b)[

==>> I would like to obtain u.pred2 such that u.pred2=g(beta, varu) for
any given beta and

varu. In other words, I need to extract the function g() used in the
lmer to predict the u random effects. Is there a way to do that?

Thank you very much.

Solange Correa,
Ph.D. student
Social Statistics
University of Southampton, UK.

************* ********************************
summary of the model fitting
**********************************************

> fit1.b
Generalized linear mixed model fit using Laplace
Formula: yij.b.true ~ x1 + x2 + (1 | area)
Data: bootsamp
AIC   BIC logLik deviance
299.1 315.5 -145.5    291.1
Random effects:
Groups Name        Variance Std.Dev.
area   (Intercept) 0.33441  0.57828
number of obs: 450, groups: area, 30

Estimated scale (compare to  1 )  0.9462897

Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.33328    0.86467  -2.698  0.00697 **
x1           0.36417    0.31776   1.146  0.25178
x2           0.15127    0.03137   4.821 1.43e-06 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Correlation of Fixed Effects:
(Intr) x1
x1 -0.125
x2 -0.960 -0.050

```